The theory of games is used to investigate several controversial issues in the literature on the balance of power. A simple model of an international system is presented as an n-person noncooperative game in extensive form, and the stability of both constant-sum and nonconstant-sum systems is examined. It is shown not only that constant-sum systems with any number of actors from two through five can be stable, but also that stability is actually promoted by conflict of interest. Contrary to much of the literature, however, there is a well-defined sense in which the most stable system is one with three actors. In each type of system, there is at least one distribution of power that leads not only to system stability but also to peace. Some of these peaceful distributions are more stable than others, and these more stable distributions are shown to be characterized by inequality rather than by equality of power. It is possible to distinguish between a bipolar and a multipolar type of stable distribution, the properties of each of which resemble, to some degree, assertions made about them in the literature. Finally, contrary to much of the recent literature on international cooperation, an increase in the ability of states to make binding agreements may actually diminish the stability of international systems.